Performance analysis of iris based recognition system at the matching score level

Over the past three years, iris based personal identification has gained considerable attention both from research groups and government organizations. Public acceptance of this biometric grew substantially too. Modern cameras used for iris acquisition are less intrusive compared to earlier iris scanning devices and public awareness of system reliability is slowly developing. A typical iris system consists of four major subsystems: (i) image acquisition, (ii) preprocessing, (iii) encoding, (iv) decision making. Most current research is focused on redesigning preprocessing and encoding techniques for iris systems. However, a framework for comprehensive analysis of iris recognition systems or a study on how various preprocessing steps influence performance of iris-based identification system does not exist. In this thesis, we propose a methodology to predict performance of a large-scale iris recognition system based on a small testing database available, using information theoretic approach.;In this work, we consider a practical setting where only matching scores are accessible for collecting data. We assume that multiple scans from the same iris are available. (Abstract shortened by UMI.)